---
license: mit
---
We performed QLoRA fine-tuning on the Baichuan2 Chat 7B model using our self-constructed mathematical reasoning dataset, resulting in its performance on GSM8K improving from 3% to 10%.